How to Use Git Branches as Low-Risk Experiments
Every branch is a hypothesis you can discard without cost.
[ essay ]
A short-lived branch is a hypothesis you can delete without a meeting.
I use branches when I am not sure the idea deserves main. mystic-bytes essay rewrites live on a branch until the word count and the voice pass both hold. If they do not, the branch dies. Nothing on main has to pretend the experiment was a plan. Merge when evidence supports. Delete when it does not. Abandoned branches are compost. They informed the path you shipped. That is not failure. That is the cheap version of finding out — cheaper than shipping a guess and reverting in public.
They let long-lived branches become second products. Integration tax hides in the merge you keep postponing until the branch is a museum of conflicts. Keep them short. Merge conflicts are decisions — choose which reality to keep, deliberately, instead of accepting whichever side the tool picked first. A conflict is a meeting you scheduled by waiting. Hold it on purpose.
Git’s gift is cheap rollback. Use it for tries that might fail. If discarding the work requires a ceremony, the branch already lived too long.
— JV · Dark Heart Labs.